Serious Flaws Found With AI's Help in the Software Behind Most Websites Set Off a Race Between Fixers and Attackers
The people who maintain PHP, the behind-the-scenes software that runs about 70% of websites, rushed out fixes today for two flaws that let a stranger take over a web server without logging in. Researchers using AI programs found the flaws, and attacks had already started before the fixes came out. Britain's cybersecurity agency is telling organizations to update now or take their sites offline.
PHP is a programming language most people have never heard of, but it runs much of the internet: about 70% of websites, including those built with popular tools like WordPress. Today its maintainers released emergency fixes for two flaws in PHP itself, which they rate as critical.
The flaws, nicknamed PHP2Shell, let an attacker who hasn't logged in send a website a specially crafted request and make the server run the attacker's own commands, or read what's in its memory, which can include passwords and keys. Because the problem is in PHP, not in any one website, it affects sites no matter how carefully they were built, and versions going back more than a decade.
AI played a central part in finding them. A security firm called Calif reported one of the flaws in August as part of a project run with OpenAI, the maker of ChatGPT, and OpenAI reported the same problem that day. Another firm, Wiz, says it independently found both using AI programs from Google and OpenAI, and built working attacks on WordPress and other popular website tools within days. "Frontier AI models substantially reduced that barrier, turning a longstanding flaw into an exploit," Wiz wrote. Wiz says one researcher did in a week what "would have taken a world-class team months of work." Early signs of one of the flaws had turned up in an outside check of PHP in 2024, but no one then could see how to attack it, and the fix stalled. The PHP Foundation, which supports the language, calls that "a structural failure."
The fixes were meant to come out tomorrow. They were rushed out today after word of the flaws leaked yesterday, and one company published its fix hours before the agreed time. Cloudflare, a company that shields websites from attacks, says it was already seeing attack traffic aimed at its customers before the fixes were public. Britain's National Cyber Security Centre is urging organizations to update every copy of PHP they run, and to take systems offline if they can't protect them quickly.
The same AI that found the flaws is helping to fix them: the PHP team says it now uses AI to check the flood of reports it receives, and a Google DeepMind security leader called frontier AI "a game-changer for cyber defense." But the PHP Foundation also says AI means it "receives far more reports than before," and Wiz's own account shows how quickly the same tools can turn a hidden weakness into a working attack. Who comes out ahead, the defenders or the attackers, is the question the next few days will start to answer.
Sources
- The PHP Foundation — Today's PHP security release, and how we're strengthening PHP securitythephp.foundation · primary
- Wiz — PHP2Shell: how Wiz found and exploited the PHP flaws with AIwiz.io
- Cloudflare — Cloudflare deploys protections against php2shell vulnerabilityblog.cloudflare.com
- UK National Cyber Security Centre — Critical security issue affecting PHPncsc.gov.uk
- BleepingComputer — PHP ships emergency patches for critical PHP2Shell flawsbleepingcomputer.com